Skip to main content
Glama
Meshy AI MCP Server
by pasie15

Meshy AI MCP Server

This is a Model Context Protocol (MCP) server that wraps the Meshy AI API. It enables MCP clients (like Claude Desktop, Cursor, Cline) to interact with Meshy's generative 3D tools directly.

Features

  • Text-to-3D: Generate 3D models from text prompts.

  • Image-to-3D: Create 3D models from reference images.

  • Multi-Image-to-3D: Create 3D models from multiple reference images.

  • Retexture: Apply new textures to existing 3D models. The text-to-texture tools are compatibility aliases for the same retexture API.

  • Text-to-Image: Generate images from text prompts.

  • Image-to-Image: Generate new images from input images.

  • Model Optimization: Remesh and optimize geometry.

  • Rigging: Auto-rig 3D characters for animation.

  • Animation: Apply animations to rigged characters.

  • Streaming: Real-time progress updates for long-running tasks.

  • Task Deletion: Delete tasks across all API categories.

Related MCP server: Meshy MCP Server

Installation

You can run the server directly using npx without installing it globally.

{
  "mcpServers": {
    "meshy-ai": {
      "command": "npx",
      "args": [
        "-y",
        "meshy-ai-mcp-server"
      ],
      "env": {
        "MESHY_API_KEY": "your_meshy_api_key_here"
      }
    }
  }
}

Option 2: Clone and Build Locally

If you want to modify the code or run it from a local source:

  1. Clone the repository:

    git clone <repository-url>
    cd meshy-ai-mcp-server
  2. Install dependencies:

    npm install
  3. Build the project:

    npm run build
  4. Configure your MCP Client:

    Add the following to your MCP client configuration (e.g., claude_desktop_config.json or VS Code settings):

    {
      "mcpServers": {
        "meshy-ai": {
          "command": "node",
          "args": [
            "/absolute/path/to/meshy-ai-mcp-server/dist/index.js"
          ],
          "env": {
            "MESHY_API_KEY": "your_meshy_api_key_here"
          }
        }
      }
    }

Configuration

You need a Meshy AI API key to use this server.

  1. Get your API key from the Meshy Dashboard.

  2. Set the MESHY_API_KEY environment variable in your MCP client configuration (as shown above).

Optional Environment Variables

  • MESHY_API_BASE: Override the API base URL (default: https://api.meshy.ai/openapi).

  • MESHY_STREAM_TIMEOUT_MS: Timeout for streaming responses in milliseconds (default: 300000 aka 5 minutes).

Troubleshooting

If your MCP client reports that the server closed during initialize, check that the client configuration passes MESHY_API_KEY into the server process. The server can start without the key so clients can inspect available tools, but Meshy API tool calls will fail until the key is configured.

Development

To run the server in development mode with auto-reloading:

# Create a .env file
echo "MESHY_API_KEY=your_key_here" > .env

# Run in dev mode
npm run dev

Available Tools

  • Text to 3D: create_text_to_3d_task, retrieve_text_to_3d_task, list_text_to_3d_tasks, stream_text_to_3d_task, delete_text_to_3d_task

  • Image to 3D: create_image_to_3d_task, retrieve_image_to_3d_task, list_image_to_3d_tasks, stream_image_to_3d_task, delete_image_to_3d_task

  • Multi-Image to 3D: create_multi_image_to_3d_task, retrieve_multi_image_to_3d_task, list_multi_image_to_3d_tasks, stream_multi_image_to_3d_task, delete_multi_image_to_3d_task

  • Retexture: create_retexture_task, retrieve_retexture_task, list_retexture_tasks, stream_retexture_task, delete_retexture_task

  • Texturing (retexture aliases): create_text_to_texture_task, retrieve_text_to_texture_task, list_text_to_texture_tasks, stream_text_to_texture_task, delete_text_to_texture_task — same retexture API, historical tool names kept for compatibility.

  • Text to Image: create_text_to_image_task, retrieve_text_to_image_task, list_text_to_image_tasks, stream_text_to_image_task, delete_text_to_image_task

  • Image to Image: create_image_to_image_task, retrieve_image_to_image_task, list_image_to_image_tasks, stream_image_to_image_task, delete_image_to_image_task

  • Remeshing: create_remesh_task, retrieve_remesh_task, list_remesh_tasks, stream_remesh_task, delete_remesh_task

  • Rigging: create_rigging_task, retrieve_rigging_task, list_rigging_tasks, stream_rigging_task, delete_rigging_task

  • Animation: create_animation_task, retrieve_animation_task, list_animation_tasks, stream_animation_task, delete_animation_task

  • Utility: get_balance

License

MIT

Available Tools

51 tools
create_animation_taskA

Apply an animation to a rigged character. rig_task_id is the ID of a completed rigging task (required). action_id is the animation action identifier from the Meshy animation library (see https://docs.meshy.ai/en/api/animation-library). post_process is optional and can contain {operation_type: 'change_fps'|'fbx2usdz'|'extract_armature', fps?: 24|25|30|60}.

Inputs: action_id (required), rig_task_id (required), post_process.

create_image_to_3d_taskB

Generate a 3D model from an input image. image_url must be a public URL or base64 data URI (.jpg/.jpeg/.png). ai_model accepts 'meshy-5', 'meshy-6', 'meshy-7', or 'latest' (Meshy 7). topology accepts 'quad' or 'triangle'. symmetry_mode accepts 'auto', 'on', or 'off'. pose_mode accepts 'a-pose', 't-pose', or '' (empty). Set should_texture=false to skip texturing and get a mesh only. origin_at accepts 'bottom' or 'center'.

Inputs: image_url (required), ai_model, topology, auto_size, +10 more.

create_image_to_image_taskA

Generate a new image based on reference images and a text prompt. ai_model is required and accepts 'nano-banana', 'nano-banana-2', 'nano-banana-pro', or 'gpt-image-2'. reference_image_urls is required: an array of 1–5 public URLs or base64 data URIs. prompt describes the transformation. aspect_ratio for the nano-banana family: '1:1', '16:9', '9:16', '4:3', or '3:4'; for gpt-image-2: '1:1', '3:2', or '2:3' (default '1:1'). Cannot combine aspect_ratio with generate_multi_view. Set generate_multi_view=true to generate multi-angle views. Set remove_background=true to return a transparent RGBA PNG.

Inputs: prompt (required), ai_model (required), reference_image_urls (required), aspect_ratio, +2 more.

create_multi_image_to_3d_taskB

Generate a 3D model from 1–4 input images. image_urls must be an array of 1–4 public URLs or base64 data URIs (.jpg/.jpeg/.png). ai_model accepts 'meshy-5', 'meshy-6', 'meshy-7', or 'latest' (Meshy 7). topology accepts 'quad' or 'triangle'. symmetry_mode accepts 'auto', 'on', or 'off'. pose_mode accepts 'a-pose', 't-pose', or '' (empty). Set should_texture=false to skip texturing and get a mesh only. origin_at accepts 'bottom' or 'center'.

Inputs: image_urls (required), ai_model, topology, auto_size, +10 more.

create_remesh_taskA

Remesh and optimize an existing 3D model. Provide either input_task_id (ID of a completed Meshy task) or model_url (public URL or data URI to a .glb/.gltf/.obj/.fbx/.stl file). topology accepts 'quad' or 'triangle'. origin_at accepts 'bottom' or 'center'. Set convert_format_only=true to only convert the format without remeshing.

Inputs: no named parameters.

create_retexture_taskA

Apply new textures to a 3D model using text or image style guidance. Provide either input_task_id (ID of a completed Meshy task) or model_url (public URL or data URI to a .glb/.gltf/.obj/.fbx/.stl file). Provide either text_style_prompt (up to 600 chars) or image_style_url to guide the texture style; these are mutually exclusive. ai_model accepts 'meshy-5', 'meshy-6', 'meshy-7', or 'latest' (Meshy 7). enable_original_uv (default true) preserves existing UV mapping. enable_pbr generates additional PBR maps (metallic, roughness, normal).

Inputs: no named parameters.

create_rigging_taskA

Auto-rig a 3D character model for animation. Provide either input_task_id (ID of a completed Meshy task) or model_url (public URL or data URI to a GLB file). height_meters is the approximate character height in meters (default 1.7, range 0.1–15). texture_image_url is an optional URL to the UV-unwrapped base color texture (PNG). Note: auto-rigging works best on humanoid characters with a clear limb/body structure.

Inputs: no named parameters.

create_text_to_3d_taskA

Generate a 3D model from a text prompt using a two-step workflow. Step 1 – set mode='preview' to create a draft mesh (requires 'prompt'). Step 2 – set mode='refine' with 'preview_task_id' pointing to a SUCCEEDED preview task to add textures. ai_model accepts 'meshy-5', 'meshy-6', 'meshy-7', or 'latest' (Meshy 7). On preview, model_type accepts 'standard', 'smart-topology', or 'lowpoly'. ultra_mode is a preview-only boolean for meshy-7/latest. art_style accepts 'realistic' or 'sculpture' (deprecated for meshy-6). topology accepts 'quad' or 'triangle'. symmetry_mode accepts 'auto', 'on', or 'off'. pose_mode accepts 'a-pose', 't-pose', or '' (empty string for none). origin_at accepts 'bottom' or 'center'.

Inputs: no named parameters.

create_text_to_image_taskB

Generate an image from a text prompt. ai_model is required and accepts 'nano-banana', 'nano-banana-2', 'nano-banana-pro', or 'gpt-image-2'. aspect_ratio for the nano-banana family: '1:1', '16:9', '9:16', '4:3', or '3:4'; for gpt-image-2: '1:1', '3:2', or '2:3' (default '1:1'). pose_mode accepts 'a-pose' or 't-pose'. Set generate_multi_view=true to generate multi-angle views (cannot be combined with aspect_ratio). Set remove_background=true to return a transparent RGBA PNG.

Inputs: prompt (required), ai_model (required), pose_mode, aspect_ratio, +2 more.

create_text_to_texture_taskB

Apply textures to a 3D model using text prompts. Compatibility alias: this tool uses Meshy's /v1/retexture API (there is no live /v1/text-to-texture endpoint). model_url is required. text_style_prompt is taken from style_prompt, or object_prompt if style_prompt is omitted. enable_original_uv and enable_pbr are forwarded to retexture.

Inputs: model_url (required), object_prompt (required), art_style, enable_pbr, +4 more.

delete_animation_taskC

Delete an animation task.

Inputs: task_id (required).

delete_image_to_3d_taskC

Delete an image-to-3d task.

Inputs: task_id (required).

delete_image_to_image_taskC

Delete an image-to-image task.

Inputs: task_id (required).

delete_multi_image_to_3d_taskC

Delete a multi-image-to-3d task.

Inputs: task_id (required).

delete_remesh_taskC

Delete a remesh task.

Inputs: task_id (required).

delete_retexture_taskC

Delete a retexture task.

Inputs: task_id (required).

delete_rigging_taskD

Delete a rigging task.

Inputs: task_id (required).

delete_text_to_3d_taskD

Delete a text-to-3d task.

Inputs: task_id (required).

delete_text_to_image_taskC

Delete a text-to-image task.

Inputs: task_id (required).

delete_text_to_texture_taskC

Delete a text-to-texture task via the /v1/retexture API.

Inputs: task_id (required).

get_balanceA

Retrieve your Meshy AI account balance.

Inputs: no named parameters.

list_animation_tasksC

List animation tasks you have previously created. sort_by accepts '+created_at' or '-created_at'.

Inputs: sort_by, page_num, page_size.

list_image_to_3d_tasksC

List previously created image-to-3d tasks. sort_by accepts '+created_at' or '-created_at'.

Inputs: sort_by, page_num, page_size.

list_image_to_image_tasksC

List previously created image-to-image tasks. sort_by accepts '+created_at' or '-created_at'.

Inputs: sort_by, page_num, page_size.

list_multi_image_to_3d_tasksC

List previously created multi-image-to-3d tasks. sort_by accepts '+created_at' or '-created_at'.

Inputs: sort_by, page_num, page_size.

list_remesh_tasksC

List previously created remesh tasks. sort_by accepts '+created_at' or '-created_at'.

Inputs: sort_by, page_num, page_size.

list_retexture_tasksC

List previously created retexture tasks. sort_by accepts '+created_at' or '-created_at'.

Inputs: sort_by, page_num, page_size.

list_rigging_tasksC

List previously created rigging tasks. sort_by accepts '+created_at' or '-created_at'.

Inputs: sort_by, page_num, page_size.

list_text_to_3d_tasksC

List previously created text-to-3d tasks. sort_by accepts '+created_at' or '-created_at'.

Inputs: sort_by, page_num, page_size.

list_text_to_image_tasksC

List previously created text-to-image tasks. sort_by accepts '+created_at' or '-created_at'.

Inputs: sort_by, page_num, page_size.

list_text_to_texture_tasksC

List previously created text-to-texture tasks via the /v1/retexture API. sort_by accepts '+created_at' or '-created_at'.

Inputs: sort_by, page_num, page_size.

retrieve_animation_taskC

Retrieve the status or result of an animation task.

Inputs: task_id (required).

retrieve_image_to_3d_taskB

Retrieve the status and result of an image-to-3d task.

Inputs: task_id (required).

retrieve_image_to_image_taskC

Retrieve the status and result of an image-to-image task.

Inputs: task_id (required).

retrieve_multi_image_to_3d_taskB

Retrieve the status and result of a multi-image-to-3d task.

Inputs: task_id (required).

retrieve_remesh_taskC

Retrieve the status and result of a remesh task.

Inputs: task_id (required).

retrieve_retexture_taskC

Retrieve the status and result of a retexture task.

Inputs: task_id (required).

retrieve_rigging_taskC

Retrieve the status of a rigging task.

Inputs: task_id (required).

retrieve_text_to_3d_taskC

Retrieve the status and result of a text-to-3d task.

Inputs: task_id (required).

retrieve_text_to_image_taskB

Retrieve the status and result of a text-to-image task.

Inputs: task_id (required).

retrieve_text_to_texture_taskC

Retrieve the status and result of a text-to-texture task. Uses the /v1/retexture API.

Inputs: task_id (required).

stream_animation_taskC

Stream updates for an animation task using server-sent events.

Inputs: task_id (required), timeout.

stream_image_to_3d_taskC

Stream updates for an image-to-3d task.

Inputs: task_id (required), timeout.

stream_image_to_image_taskC

Stream updates for an image-to-image task.

Inputs: task_id (required), timeout.

stream_multi_image_to_3d_taskD

Stream updates for a multi-image-to-3d task.

Inputs: task_id (required), timeout.

stream_remesh_taskC

Stream updates for a remesh task.

Inputs: task_id (required), timeout.

stream_retexture_taskC

Stream updates for a retexture task.

Inputs: task_id (required), timeout.

stream_rigging_taskC

Stream updates for a rigging task until it completes.

Inputs: task_id (required), timeout.

stream_text_to_3d_taskC

Stream updates for a text-to-3d task until it completes.

Inputs: task_id (required), timeout.

stream_text_to_image_taskC

Stream updates for a text-to-image task.

Inputs: task_id (required), timeout.

stream_text_to_texture_taskC

Stream updates for a text-to-texture task using the /v1/retexture API.

Inputs: task_id (required), timeout.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    C2.8/5.0

    Scored across 51 tools

    Disambiguation3/5

    Most tools are distinguishable by action prefix and modality, so create_text_to_3d_task vs create_image_to_3d_task vs create_remesh_task are reasonably clear. However, the text_to_texture task family duplicates the retexture task family using the same underlying /v1/retexture API, creating real misselection risk across five pairs of tools.

    Naming Consistency5/5

    The task tools consistently follow a snake_case action_resource pattern: create_, retrieve_, list_, stream_, and delete_ combined with the task type. The only deviation is get_balance, which is a unique account operation and does not disrupt overall predictability.

    Tool Count2/5

    At 51 tools, the surface is far larger than a practical agent-facing set, with five similar operations repeated across ten task families. Some breadth is justified by Meshy's multiple modalities, but the duplicated text_to_texture/retexture family and per-task stream/list/delete tools make it excessively heavy.

    Completeness5/5

    Each task family has create, retrieve, list, stream, and delete coverage, plus balance retrieval, so the lifecycle is essentially complete. No obvious domain gap prevents an agent from running generation, refinement, rigging, animation, and cleanup workflows.

    Related MCP Connectors

    Related MCP Servers